After a long wait to replenish my thermal epoxy supply (there's
still a kit lost in the mail somewhere), I set out to make a full size version
to see what i really had my mitts into. This one is a little different in
construction, but works the same on a larger scale. 4x to be exact. I expected
this one to do fairly well.
The blue lines show the flow path of the water
Upon completion of the thing, I stuck it in my test rig and
proceeded to water down my GeForce MX. I build a lot of stuff like this and
have had pretty good luck with stuff not leaking - especially with "Marine
Goop". Apparently I hadn't done as good of job gluing it as I had thought,
so I yanked it out and carefully removed the bottom and tried to glue it again
with a little more Goop. I got wise and tested it first this time , all seemed
well until I got it installed and then the other side started leaking - It
wasn't the sprinkler it was the first time , but leaking none the less. Something
was definitely amiss here, as far as I could tell - the parts fit too closely
together to use urethane and it couldn't dry properly. I should have used
an epoxy, but was afraid the excess would have been wicked into the foam.
I had to do something quick or this whole experiment was gonna
end up a disaster. I found a can of urethane floor covering, pried the
lid off and submerged the block in it , the leaks were only pinholes so I
sucked in a little vacuum to draw the urethane into the leaks. (That's
when the wife came out and basically saw what looked like me,after being sick
as a dog all week, drinking from a varnish can with a straw, funny thing is
she just shook her head and went back inside :)
Once this stuff dried I tested again.. I HAD TO MAKE IT WORK!..
Fortunately the overclocking gods were smiling on me and no leaks! I
slapped it back into the test rig and prayed once again that we hadn't killed
anything with the water, to my relief it whirred and beeped to life like usual
(Iwill's kick @ss) and I was off and running, into windows!
A quick peek at the MBM says 38c - that's about what I get
from a good copper block at idle! I fired AHL and went in for a couple
rounds. After 40 minutes fraggin it was holding steady at 41c. Exactly the
same as a copper water block on this system.
So lets get a look at some results here..
|
Cpu Speed
|
Cpu Voltage
|
Watts
|
Ambient Temp
|
Cpu Temp
|
C/W
|
|
800mhz
|
1.60 (.65 in BIOS)
|
34.1
|
31c
|
54c
|
0.67
|
|
1000mhz
|
1.85 (0.9v in BIOS)
|
52.6
|
31c
|
41c
|
0.19
|
The Basic Definition of C/W is how many degrees
C above ambient the cpu will be per watt by the cpu. So multiply cpu wattage
by c/w to get expected temp.
This experiment reinforces my belief that if we are gonna
get any closer to the water temperature we are gonna have to eliminate
the metal and go direct water to die. (which we have been working on
for nearly a year :) Although not directly comparible (different pump set
up, different temperature probes) Spode's Sunny
D cooler got a C/W of 0.13.
